Cardiovascular disease is one of the most serious diseases that harm human life, the standard 12-lead Electrocardiogram (ECG) has been widely used to diagnose a variety of cardiac abnormalities, and predict cardiovascular morbidity and mortality for its simplicity and non-invasive property. In this study, we present an end-to-end deep residual neural network with different kernel sizes to identify the 12-lead ECG with different durations. Firstly, each ECG record is divided into segments with a fixed length of 10 s, and the total number of segments for an ECG record is 6. A novel shared-weight deep convolutional neural network is proposed to reduce the number of parameters while allowing for more robust feature detection. Secondly, multi-scale convolutional neural networks are used to extract features at different scales and frequencies, leading to superior feature representation. Finally, the Squeeze and Excitation (SE) block is integrated into standard ResNet architectures (SE-ResNet) to adaptively recalibrate channel-wise feature responses by explicitly modelling interdependencies between channels. The approach has been validated against the 2020 Physionet/CinC Challenge data set, obtaining a F2 score of 0.798, G2 score of 0.604 and Geometric Mean of 0.694 on the hidden test set, respectively. Results suggest that the proposed method achieves competitive performance in multi-label ECGs classification.